An adaptive error correction scheme for computer memory system

Many of the modern computer memories contain single-error correction capability in order to enhance reliability. In a large scale memory, an even more powerful error correction code may be desirable. In particular, a double-error correction capability can reduce the maintenance cost significantly, while keeping the unscheduled system interruptions within tolerable limits. Since most faults are effectively masked and logged out, the permanent failures can be replaced at the time of scheduled maintenance, thus leaving the user unaffected. The cost and complexity of the known double error correcting code, however, seems to outweigh the advantages. The long decoding time and large amount of redundancy in double error correction cannot be justified in every fetch instruction for the sake of correcting an occasional double error.